Five reproducible benchmarks. One trust substrate.
Each POC opens with a scenario a platform team will recognize, walks through what SolvNum changes, and ends with captured benchmark numbers plus a SHA-256 you can re-derive on your own hardware in five minutes with pip install numpy.
The POC suite
From ML decisions to creator payouts
Each POC addresses a distinct platform workload where arithmetic reproducibility is a regulatory or business requirement. Together they cover the full stack from model output to economic consequence.
AI Decision-Layer Reproducibility
1,000,000 ML decisions. SolvNum produces 0 receipt drifts across platforms. Float64 produces ~32,000. EU AI Act Article 12 ready.
SolvSRK Dynamics Under Noise
100% survival on stiff contact (k=10⁷) and VdPol μ=1000 under noise. RK4 at 2 kHz: 0%. RK45 at σ=1.0: 0%.
Ads Attribution Reconciliation
32/32 receipts identical across Windows and Linux on 6.1M real Criteo journeys. Float64 drifts on the same machine under BLAS swap.
Creator Payout Verifiable Workflow
Bit-identical payouts across Python (Win/Linux) and JavaScript BigInt. Browser-verifiable. $0.013 drift eliminated.
LLM Watermark Detection Reproducibility
Both SolvNum detectors match cross-platform. Float64 weighted detector drifts — different receipt hash, different borderline decisions.
Want the receipts on your own pipeline?
Run these benchmarks against your real decision pipeline.
A bounded-scope reproducibility audit against a sanitized snapshot of your real pipeline. Two weeks, $25K, fully credited. No production integration, no data leaving your premises. Every claim above traces back to a script you can run locally.